all_results.json

{
    "epoch": 0.9993060374739764,
    "eval_logits/chosen": -0.21596238017082214,
    "eval_logits/rejected": 0.9912689924240112,
    "eval_logps/chosen": -579.5609130859375,
    "eval_logps/rejected": -885.7299194335938,
    "eval_loss": 0.2786034047603607,
    "eval_rewards/accuracies": 0.8806390762329102,
    "eval_rewards/chosen": -1.94976806640625,
    "eval_rewards/margins": 2.806307792663574,
    "eval_rewards/rejected": -4.756075859069824,
    "eval_runtime": 387.8423,
    "eval_samples": 8491,
    "eval_samples_per_second": 21.893,
    "eval_steps_per_second": 0.343,
    "total_flos": 0.0,
    "train_loss": 0.1315754380491045,
    "train_runtime": 20812.0317,
    "train_samples": 184443,
    "train_samples_per_second": 8.862,
    "train_steps_per_second": 0.035
}